REST API
Tenant-scoped JSON over HTTP for dashboards, internal tools, and
integrations. The full data model, read-only.
MCP Server
A Model Context Protocol server so AI assistants can read guest context
with the same API keys and tenant isolation.
How it fits together
Independent connector workers (Mews, Square, …) pull from each source and publish canonical entities onto a queue. A single ingest worker writes them — idempotently — into Postgres as both the current canonical state and an append-only event log. An identity-resolution pipeline merges source records into master profiles. The API worker reads from that store. Nothing is forgotten; everything is replayable.Core principles
Tenant-scoped by construction
Tenant-scoped by construction
Every request resolves its scope from the bearer credential — the whole
property group by default, one property when the credential is pinned or
the request drills down. The tenant is never trusted from the caller, so a
token can only ever read its own data.
Read-only canonical data
Read-only canonical data
The API never writes canonical guest data — that only enters through the
connectors and the ingest pipeline. Integration endpoints write connection
metadata only, and are excluded from API-key access entirely.
One guest, resolved
One guest, resolved
Identity resolution merges each guest’s records across source systems and
properties into a single master profile — a golden record with
per-attribute provenance.
Connect once, map per property
Connect once, map per property
An integration is authenticated once per property group as a connection
(one per provider). Each property is then mapped to a specific source
entity (a Mews EnterpriseId, a Square location) under that connection.
Next steps
Quickstart
Generate an API key and make your first request in two minutes.
Authentication
Credential types, tenant scoping, and access scopes.
Data model
Every entity the API exposes and how they relate.
Browse the API
Endpoint reference with an interactive playground.